Last active
March 24, 2019 01:04
-
-
Save fehiepsi/5ef8e09e61604f10607380467eb82006 to your computer and use it in GitHub Desktop.
Re: normalization: In the current design, Gaussian
s are essentially numerically stable representations of strictly concave quadratic polynomials, i.e. they are intentionally non-normalized. Thus the log(2 pi)
and log det
terms appear only when reducing via .logaddexp()
. This is a bit tricky since we often reduce only a subset of dims, and only that subset get the log(2 pi)
terms. I don't yet know how to compute the log det
terms of the partial reduction.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
This is great! We should add it to funsor/derivations/gaussian.ipynb 😄